[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Feature request
From: |
Torbjorn Granlund |
Subject: |
Re: Feature request |
Date: |
06 Oct 2002 23:00:47 +0200 |
User-agent: |
Gnus/5.09 (Gnus v5.9.0) Emacs/21.2 |
Garrett Kajmowicz <address@hidden> writes:
I would like to see the ability to do floating point exponentials
with the mpf class. As it stands I now have to convert to double
precision, exponentiate and convert back.
Sorry, no such thing in mpf. As Paul suggested, you might want to
look at mpfr (--enable-mpfr).
I would also like to ask about precision. I am doing some work (all in
mpf_class using c++) for matrix solving. Anyways, when I do a few matrix
operations which usually result in something to the effect of 1/3 -1/3 the
answer should be zero. I am aware of the limitations of floating point
numbers, but for some reason no matter how much I increased the precision of
the mpf_class objects, the result was always the same, roughly 1e-24, using
the cout << temp function to output the value. Do you know why the value
will not drop down further (say 1e-50) when I increate to 16k bits of
precision?
This sounds a bit like a bug in mpf. Please investigate and report.
As well, is there a possability of adding a function which will
increase/decrease an mpf number by the minimal step size possible given the
current precision? This might temporarily provide an answer to the above
problem.
The minimal step size is 32 bits for a 32-bit computer and 64 bits for
a 64-bit computer.
--
Torbjörn